LETTER doi:10.

1038/nature10377

An extremely primitive star in the Galactic halo

Elisabetta Caffau1,2, Piercarlo Bonifacio2, Patrick François2,3, Luca Sbordone1,2,4, Lorenzo Monaco5, Monique Spite2, François Spite2,
Hans-G. Ludwig1,2, Roger Cayrel2, Simone Zaggia6, François Hammer2, Sofia Randich7, Paolo Molaro8 & Vanessa Hill9

The early Universe had a chemical composition consisting of hydro-
gen, helium and traces of lithium1; almost all other elements were
subsequently created in stars and supernovae. The mass fraction of
elements more massive than helium, Z, is known as ‘metallicity’. A
number of very metal-poor stars has been found2,3, some of which
have a low iron abundance but are rich in carbon, nitrogen and
oxygen4–6. For theoretical reasons7,8 and because of an observed
absence of stars with Z , 1.5 3 1025, it has been suggested that
low-mass stars cannot form from the primitive interstellar medium
until it has been enriched above a critical value of Z, estimated to lie
in the range 1.5 3 1028 to 1.5 3 1026 (ref. 8), although competing
theories claiming the contrary do exist9. (We use ‘low-mass’ here to
mean a stellar mass of less than 0.8 solar masses, the stars that
survive to the present day.) Here we report the chemical composi-
tion of a star in the Galactic halo with a very low Z (# 6.9 3 1027,
which is 4.5 3 1025 times that of the Sun10) and a chemical pattern
typical of classical extremely metal-poor stars2,3—that is, without
enrichment of carbon, nitrogen and oxygen. This shows that low-
mass stars can be formed at very low metallicity, that is, below the
critical value of Z. Lithium is not detected, suggesting a low-
metallicity extension of the previously observed trend in lithium
depletion11. Such lithium depletion implies that the stellar material
must have experienced temperatures above two million kelvin in its
history, given that this is necessary to destroy lithium.
The Galactic halo star SDSS J1029151172927, the object of this
Letter, has been observed with the X-Shooter12 and UVES13 spectro-
the primitive interstellar medium through the fine structure lines of
ionized carbon and neutral oxygen. A suitable combination of the
carbon and oxygen abundances is called the transition discriminant16
(D 5 log10(10[C/H] 1 0.3 3 10[O/H])), and low-mass star formation is
believed to occur only if D $ 23.5. From the abundances in Table 1
and the assumption [O/Fe] 5 10.6 we have D # 24.2 for SDSS
J1029151172927, which places it in the ‘forbidden zone’ of the theory.
If, instead of taking the upper limit on the carbon abundance, we
assume that the carbon abundance (derived from the three-dimen-
sional (3D) analysis) scales with the iron abundance, as found in other
metal-poor stars3, we have D # 24.4. Our measurement cannot rule
out the above-mentioned theoretical scenario7,16, but it strongly sup-
ports the idea that, at least in some cases, low-mass stars can also form
at lower carbon and oxygen abundances than the current estimates for
the critical values.
The complete absence of the neutral lithium (Li I) resonance doublet
at 670.7 nm, both in UVES and X-Shooter spectra, is remarkable. In
2.0
X–Shooter
1.5
Ca II K
Normalized flux

Ca II H
graphs at the Very Large Telescope (VLT), operated by the European UVES
1.0
Southern Observatory in Chile. Its properties are as follows: right
ascension, 10 h 29 min 15.15 s, declination, 117u 299 280 at equinox
Ca II K Ca II H
2000; g band magnitude 16.92, (g 2 z) 5 0.59 mag, and, after correc- 1.0 1.0
0.5
tion for interstellar reddening, (g–z)0 5 0.53 mag. A portion of the 0.8 0.8
spectra in the region of the Ca II K line is shown in Fig. 1. We have 0.6 0.6
computed and used theoretical model atmospheres and spectrum syn- 0.4 0.4
0.0
thesis techniques to derive the chemical abundances provided in 0.2 IS 0.2
IS
Table 1. The chemical signatures are consistent with metal production H H
393.2 393.3 393.4 396.6 396.8 397.0 397.2
by ordinary core-collapse supernovae14. The derived abundances,
coupled with the upper limits on carbon and nitrogen, imply 388 390 392 394 396 398
Z # 6.9 3 1027. This number takes into account the typical ‘excess’ Wavelength (nm)
of the a-element oxygen, [O/Fe] 5 10.6. (Here [A/B] 5 log(NA/ Figure 1 | Observed spectra of SDSS J1029151172927. The spectral region
NB) 2 log(NA/NB)[ for the number N of atoms of elements A and of the Ca II H and K lines is shown (solid lines), compared to synthetic spectra
B, and subscript [ indicates the solar value.) Our analysis has been (long dashed lines) computed with a global metallicity of Z 5 1.1 3 1026 and
performed assuming local thermodynamic equilibrium (LTE); however, solar proportions of all elements, except for a-elements that are enhanced by
further work is necessary to assess the role of departures from LTE, 0.4 dex over iron. Main figure: top trace, the X-Shooter spectrum (shifted
especially for molecules. The estimate of non-LTE effects on magnesium15 vertically by one unit for clarity); bottom trace, the UVES spectrum. The spectra
have been normalized to 1 in the continuum. Insets, magnified views of the Ca II
is about 10.4 dex, which translates in a change of 10.2 3 1027 in Z.
K line (left) and the Ca II H line (right). The absorption due to interstellar gas is
It has been suggested that the primary discriminants between the clearly detectable both in K and H Ca II lines (labelled as ‘IS’ in the figure), and
formation of only massive stars (as in stellar population III) and of two hydrogen lines, H8 and He (labelled as ‘H’ in the figure), are visible. The
both massive and low-mass stars (as in stellar populations II and I) measured radial velocity is 234.5 6 1.0 km s21.We computed a Galactic orbit
are the abundances of carbon and oxygen7. In this scenario, these from the kinematic data and a distance of 1.27 6 0.15 kpc, estimated from the
elements can provide efficient cooling of the protostellar clouds in photometry, confirming that the star belongs to the Galactic halo.
1
Zentrum für Astronomie der Universität Heidelberg, Landessternwarte, Königstuhl 12, 69117 Heidelberg, Germany. 2GEPI, Observatoire de Paris, CNRS, Université Paris Diderot, Place Jules Janssen,
92190 Meudon, France. 3UPJV, Université de Picardie Jules Verne, 33 Rue St Leu, F-80080 Amiens, France. 4Max-Planck Institut für Astrophysik, Karl-Schwarzschild-Strasse 1, 85741 Garching, Germany.
5
European Southern Observatory, Casilla 19001, Santiago, Chile. 6Istituto Nazionale di Astrofisica, Osservatorio Astronomico di Padova Vicolo dell’Osservatorio 5, 35122 Padova, Italy. 7Istituto Nazionale di
Astrofisica, Osservatorio Astrofisico di Arcetri, Largo E. Fermi 5, 50125 Firenze, Italy. 8Istituto Nazionale di Astrofisica, Osservatorio Astronomico di Trieste, Via Tiepolo 11, 34143 Trieste, Italy. 9Université de
Nice-Sophia Antipolis, Observatoire de la Côte d’Azur, CNRS, Laboratoire Cassiopée, Boulevard de l’Observatoire, 06300 Nice, France.

1 S E P T E M B E R 2 0 1 1 | VO L 4 7 7 | N AT U R E | 6 7
©2011 Macmillan Publishers Limited. All rights reserved
 RESEARCH LETTER

Table 1 | Abundances in SDSS J1029151172927 a 3

Element A(X), [X/H], [X/Fe], [X/H], Number of A(X)[
3D 3D 3D 1D lines

C #4.2 #24.3 #10.7 #23.8 G band 8.50

N #3.1 #24.8 #10.2 #24.1 NH band 7.86 2
Mg I 2.95 24.59 6 0.10 10.40 24.68 6 0.08 4 7.54

A(Li)
Si I 3.25 24.27 6 0.10 10.72 24.27 6 0.10 1 7.52
Ca I 1.53 24.80 6 0.10 10.19 24.72 6 0.10 1 6.33
Ca II 1.48 24.85 6 0.11 10.14 24.71 6 0.11 3 6.33 SDSS J102915+172927
Ti II 0.14 24.76 6 0.11 10.23 24.75 6 0.11 6 4.90 1
Fe I 2.53 24.99 6 0.12 10.00 24.73 6 0.13 44 7.52 HE 1327–2326
Ni I 1.35 24.88 6 0.11 10.11 24.55 6 0.14 10 6.23
Sr II #22.28 #25.2 #20.21 #25.1 1 2.92
These elemental abundances are derived from our UVES spectra. The last column provides the adopted
0
solar abundances on the scale A(X) 5 log10(X/H) 1 12. The atmospheric parameters adopted are –6 –5 –4 –3
effective temperature Teff 5 5,811 K, log g 5 4.0 (where g is acceleration due to gravity, in cm s22, at the
[Fe/H]
surface), and microturbulent velocity 1.5 km s21. Teff was derived from the (g 2 z)0 colour–Teff
calibration26. The combination of photometric and reddening uncertainties causes an uncertainty on
Teff of 100 K, the corresponding uncertainty on [Fe/H] is 0.06 dex. We cross-checked the values of Teff b 3
with a fit of the Ha wings which provided the same effective temperature within 10 K. The surface gravity
has been fixed from the Balmer jump, as measured by the (u 2 g) colour. Other gravity indicators, such
as the calcium ionization equilibrium and the wings of higher-order Balmer lines, are consistent with
CS 22876–32 A
this choice. The uncertainty on the surface gravity is about 0.2 dex, and we can robustly exclude a
surface gravity log g 5 3.0 or lower, thus excluding that the star could be on the horizontal branch. We
2 CS 22876–32 B
computed synthetic spectra with the SYNTHE code27 and a one-dimensional (1D) model atmosphere
was computed with the ATLAS 9 code27. These synthetic spectra were used to perform line-profile fitting

A(Li)
G 139–8
for all the measurable features. The 3D corrections were computed using a 3D model atmosphere from
the CIFIST grid28 with Teff 5 5,850 K, log g 5 4.0, and metallicity 2.7 3 1025. We were able to measure
SDSS J102915+172927 G 186–26
the abundances of only some a-elements (Mg, Ca, Si, Ti) and two iron peak elements (Fe and Ni). The G 122–69
derived iron abundance is [Fe/H] 5 24.99 (see Table 1; the 3D-corrected abundances in columns 2, 3 1
and 4 should be used, the 1D abundances in column 5 are given for reference only). The a-elements are HE 1327–2326
slightly enhanced relative to iron, [Mg/Fe] 5 10.4. The Sr II line at 407.8 nm is not convincingly
detected, giving an upper limit [Sr/Fe] # 20.21, which is compatible with the general pattern of low [Sr/
Fe] found in extremely metal-poor stars29. The upper limits on carbon and nitrogen are derived by fitting
the molecular bands of CH (G band) and NH, at 430 nm and 336 nm, respectively. Unfortunately no
0
measurement of oxygen is possible in the available spectral range, neither from atomic nor from –4 –3.5 –3 –2.5 –2
molecular lines, but there is no reason to suspect that a star not enhanced in Mg, C and N should be
[C/H]
over-abundant in oxygen.

Figure 2 | Lithium abundance of SDSS J1029151172927 compared to that

fact most of the ‘warm’ (effective temperature Teff . 5,700 K) metal-
poor dwarf stars display a constant abundance of Li, the so-called Spite
plateau11,17. From the signal-to-noise ratio in the UVES spectrum of
SDSS J1029151172927, we derive an upper limit for the Li abundance,
A(Li) , 1.1 (at 5s). In Fig. 2 we show the Spite plateau as a function of
the carbon abundance, as well as a function of the iron abundance,
which we use in turn as a proxy for Z. The sample of stars is composed
of those with a normal carbon abundance3,18–20 and the carbon-rich,
iron-poor subgiant HE 132722326 (ref. 4). The pictures emerging
from Fig. 2a and b show the same morphology, with the exception of
star HE 132722326, which has [Fe/H] lower than all the others, but has
[C/H] comparable to many other stars in the sample. It is noteworthy
that the only two unevolved stars with [Fe/H] , 24.5 have no detect-
able Li.
The most straightforward interpretation of the Spite plateau is that
the Li observed in the plateau stars is the Li produced in the Big Bang17.
The theoretical primordial Li abundance1 is a factor of 2–3 larger than
the value observed on the Spite plateau. A number of explanations of
this discrepancy have been proposed, which range from stellar phe-
nomena, such as atomic diffusion21, to new physics leading to a dif-
ferent Big Bang nucleosynthesis22. Our upper limit implies that the Li
abundance of SDSS J1029151172927 is far below the value of the Spite
plateau. At extremely low metallicities, the Spite plateau displays a
‘meltdown’11, that is, an increased scatter and a lower mean Li abund-
ance. This meltdown is clearly seen in the two components of the
extremely metal-poor binary system CS 22876-32, which show a dif-
ferent Li content19. The primary is on the Spite plateau, whereas the
secondary is below it, at A(Li) 5 1.8. The reasons for this meltdown are
not understood. It has been suggested11 that a Li depletion mechanism,
whose efficiency depends on metallicity and temperature, could
explain the observations. If this were the case, the Li abundance in
SDSS J1029151172927 would result from efficient Li depletion due to
a combination of extremely low metallicity and relatively low temper-
ature. For completeness, we mention that there are a small number of
known stars which have a metallicity and effective temperature similar
to that of other stars on the Spite plateau, but where the Li doublet is
of other metal-poor stars. a, b, The Spite plateau is shown as a function of iron
abundance, [Fe/H] (a), and of carbon abundance, [C/H] (b). We use carbon
and iron in turn as proxies of the metallicity Z. The upper limits for SDSS
J1029151172927 are from the present work. The other filled black circles are
from refs 3 and 18, the upper limit for HE 132722326 is from ref. 4. The other
open black circles are the upper limits and are mentioned below. The Li
measurements for the binary system CS 22876–32 are from ref. 19, the upper
limits for the three well known Li-depleted dwarfs, G 122-69, G 139-8 and G
186-26, are from ref. 20. For these three last stars, as well as for CS 22876–32, we
have no measurement of the C abundances, and we therefore assumed that C
scales with iron as in the rest of the sample3. The precise placement of these stars
along the abscissa in this diagram is of no consequence for the present
discussion.
not detected. The fact that such stars are found for different values of
[Fe/H] and [C/H] suggests that Li-depletion is independent of either. It
has been suggested that Li-depleted stars could have a common origin
with blue stragglers23, an interpretation that has been reinforced by the
discovery that these stars are also depleted in beryllium24.
Stars similar to SDSS J1029151172927 are probably not very rare.
Only 30% of the whole SDSS survey area was accessible to our VLT
observations. We identified 2,899 potentially extreme stars with metal-
licity less than Z # 1.1 3 1025 in Data Release 725. Among those
observable with the VLT, we performed a subjective selection of the
most promising candidates; of these, we observed six in our X-Shooter
programme, resulting in one detection. Depending on the subjective
bias we attribute to the last selection step, we expect 5–50 stars of
similar or even lower metallicity than SDSS J1029151172927 to be
found among the candidates accessible from the VLT, and many more
in the whole SDSS sample.
Received 31 March; accepted 15 July 2011.
1. Iocco, F., Mangano, G., Miele, G., Pisanti, O. & Serpico, P. D. Primordial
nucleosynthesis: from precision cosmology to fundamental physics. Phys. Rep.
472, 1–76 (2009).
2. Cayrel, R. et al. First stars V — Abundance patterns from C to Zn and supernova
yields in the early Galaxy. Astron. Astrophys. 416, 1117–1138 (2004).

6 8 | N AT U R E | VO L 4 7 7 | 1 S E P T E M B E R 2 0 1 1
©2011 Macmillan Publishers Limited. All rights reserved

3. Bonifacio, P. et al. First stars XII. Abundances in extremely metal-poor turnoff stars,
and comparison with the giants. Astron. Astrophys. 501, 519–530 (2009).
4. Frebel, A., Collet, R., Eriksson, K., Christlieb, N. & Aoki, W. HE 1327–2326, an
unevolved star with [Fe/H],25.0. II. New 3D–1D corrected abundances from a
Very Large Telescope UVES spectrum. Astrophys. J. 684, 588–602 (2008).
5. Christlieb, N. et al. A stellar relic from the early Milky Way. Nature 419, 904–906
(2002).
6. Norris, J. E. et al. HE 0557–4840: ultra-metal-poor and carbon-rich. Astrophys. J.
670, 774–788 (2007).
7. Bromm, V. & Loeb, A. The formation of the first low-mass stars from gas with low
carbon and oxygen abundances. Nature 425, 812–814 (2003).
8. Schneider, R., Ferrara, A., Salvaterra, R., Omukai, K. & Bromm, V. Low-mass relics of
early star formation. Nature 422, 869–871 (2003).
9. Nakamura, F. & Umemura, M. On the initial mass function of Population III stars.
Astrophys. J. 548, 19–32 (2001).
10. Caffau, E., Ludwig, H.-G., Steffen, M., Freytag, B. & Bonifacio, P. Solar chemical
abundances determined with a CO5BOLD 3D model atmosphere. Sol. Phys. 268,
255–269 (2011).
11. Sbordone, L. et al. The metal-poor end of the Spite plateau. I. Stellar parameters,
metallicities, and lithium abundances. Astron. Astrophys. 522, A26 (2010).
12. D’Odorico, S. et al. X-shooter UV- to K-band intermediate-resolution high-
efficiency spectrograph for the VLT: status report at the final design review. Proc.
SPIE 6269, 626933 (2006).
13. Dekker, H., D’Odorico, S., Kaufer, A., Delabre, B. & Kotzlowski, H. Design,
construction, and performance of UVES, the echelle spectrograph for the UT2
Kueyen Telescope at the ESO Paranal. Proc. SPIE 4008, 534–545 (2000).
14. Chieffi, A. & Limongi, M. The explosive yields produced by the first generation of
core collapse supernovae and the chemical composition of extremely metal poor
stars. Astrophys. J. 577, 281–294 (2002).
15. Andrievsky, S. M. et al. Non-LTE abundances of Mg and K in extremely metal-poor
stars and the evolution of [O/Mg], [Na/Mg], [Al/Mg], and [K/Mg] in the Milky Way.
Astron. Astrophys. 509, A88 (2010).
16. Frebel, A., Johnson, J. L. & Bromm, V. Probing the formation of the first low-mass
stars with stellar archaeology. Mon. Not. R. Astron. Soc. 380, L40–L44 (2007).
17. Spite, M. & Spite, F. Lithium abundance at the formation of the Galaxy. Nature 297,
483–485 (1982).
18. Bonifacio, P. et al. First stars VII — Lithium in extremely metal poor dwarfs. Astron.
Astrophys. 462, 851–864 (2007).
19. González Hernández, J. I. et al. First stars XI. Chemical composition of the extremely
metal-poor dwarfs in the binary CS 22876–032. Astron. Astrophys. 480, 233–246
(2008).
20. Norris, J. E., Ryan, S. G., Beers, T. C. & Deliyannis, C. P. Extremely metal-poor stars. III.
The Li-depleted main-sequence turnoff dwarfs. Astrophys. J. 485, 370–379 (1997).
©2011 Macmillan Publishers Limited. All rights reserved
LETTER RESEARCH
21. Richard, O., Michaud, G. & Richer, J. Implications of WMAP observations on Li
abundance and stellar evolution models. Astrophys. J. 619, 538–548 (2005).
22. Jedamzik, K. & Pospelov, M. Big Bang nucleosynthesis and particle dark matter. N.
J. Phys. 11, 105028 (2009).
23. Ryan, S. G., Gregory, S. G., Kolb, U., Beers, T. C. & Kajino, T. Rapid rotation of ultra-Li-
depleted halo stars and their association with blue stragglers. Astrophys. J. 571,
501–511 (2002).
24. Boesgaard, A. M. Beryllium in ultra-lithium-deficient halo stars: the blue straggler
connection. Astrophys. J. 667, 1196–1205 (2007).
25. Abazajian, K. N. et al. The seventh data release of the Sloan Digital Sky Survey.
Astrophys. J. 182 (Suppl.), 543–558 (2009).
26. Ludwig, H.-G. et al. Extremely metal-poor stars from the SDSS. Phys. Scripta T 133,
014037 (2008).
27. Kurucz, R. L. ATLAS12, SYNTHE, ATLAS9, WIDTH9, et cetera. Mem. Soc. Astron. Ital.,
Suppl. 8, 14–24 (2005).
28. Ludwig, H.-G. et al. The CIFIST 3D model atmosphere grid. Mem. Soc. Astron. Ital.
80, 711–714 (2009).
29. François, P. et al. First stars. VIII. Enrichment of the neutron-capture elements in the
early Galaxy. Astron. Astrophys. 476, 935–950 (2007).
Acknowledgements The spectra were secured through X-Shooter Guaranteed Time
Observations (GTOs) and ESO Director’s Discretionary Time at the ESO VLT Kueyen
8.2-m telescope. E.C. is a Gliese fellow.
Author Contributions E.C. developed the code for the analysis of the SDSS spectra,
selected the targets for high resolution follow-up, performed the chemical analysis of
X-Shooter and UVES spectra and was mainly responsible for writing the paper. P.B.
supervised the project and was the Principal Investigator of the ESO proposals. P.F.
reduced the X-Shooter data and cross-checked the chemical analysis. L.S. reduced the
UVES data. L.M. performed the X-Shooter observations. M.S. and F.S. cross-checked the
chemical analysis and contributed to writing the paper. H.-G.L. provided the codes for
3D hydrodynamical simulations and spectral synthesis. R.C. was the main inspirer of
this project. S.Z. was responsible for the interpretation of the kinematical data. F.H. and
S.R. were respectively the French and Italian Principal Investigators of X-Shooter, who
were granted the GTO and decided to invest it in this project. P.M., V.H. and all the other
authors contributed to the astrophysical interpretation and to the final version of the
paper.
Author Information Reprints and permissions information is available at
www.nature.com/reprints. The authors declare no competing financial interests.
Readers are welcome to comment on the online version of this article at
www.nature.com/nature. Correspondence and requests for materials should be
addressed to E.C. (Elisabetta.Caffau@obspm.fr).
1 S E P T E M B E R 2 0 1 1 | VO L 4 7 7 | N AT U R E | 6 9